How to apply "harmoniously" with AI?A number of technologies debut 2022 World Artificial Intelligence Conference

Cover news reporter Zhang Yuexi

With the deepening of artificial intelligence applications, its own technical defects, the problems of decision -making bias, and the safety of the use of the trust have aroused the crisis of trust, which can be trusted to become the focus of attention. The construction of trusted artificial intelligence and solving the trust in artificial intelligence applications has become an inevitable choice to relieve and eliminate these concerns.

On September 3, a number of important technologies were at the 2022 World Artificial Intelligence Conference.

Hidden open platform is launched globally

Support low -code calling privacy computing technology

2022 The World Artificial Intelligence Conference (WAIC) Trusting Privacy Computing Summit Forum, Ant Group released a trusted privacy computing "hidden open platform", which is open to global users. Explore the application scenario of privacy computing.

In the circulation of large -scale data of artificial intelligence, guarding data security and user privacy are the cornerstones of digitalization. Trusting privacy computing is considered a supportive technology to solve this problem. How to promote privacy computing product functions to more scenarios to more scenarios and wider ecology in the way of simple and easy to deploy, low -cost, has always been a challenge facing the industry.

In this context, after the open source "hidden language" trusted privacy computing framework in July, the hidden open platform was released. Users can choose the "white screen" mode without calling and developing code, and directly use the product function. They can also choose the "black screen" mode of calling the code and flexible assembly function. The hidden language frame and hidden open platform can be understood as a flexible combination code "material package", and the latter provides a "bag in" type of "bags in".

Privacy Cooperation Platform FAIR Upgrade

Extra -level data security matching ability is increased by more than 10 times

As we all know, in financial risk control, joint marketing, medical and health and other scenarios, frequent and large number of multi -party data computing tasks are often involved, the results can be calculated quickly, accurate, and safely, and directly affect the cooperation efficiency of business.

The reporter learned that at this conference, the privacy collaboration platform FAIR announced the overall architecture upgrade. FAIR is compatible with full type of privacy computing selection. At present, it has achieved the connection between the open source framework such as hidden language, Fate and other privacy computing systems such as Moses. This means that users can call multiple privacy components such as multi -party security computing, federal learning, and trusted execution environment on a single platform to meet the needs of multiple participants' data security, efficient collaborative computing and circulation scenarios.

Specifically, it can solve problems such as "available and invisible" in the data circulation process, and problems and authenticity of data ownership and authenticity between multi -party collaboration. , Registration and transaction consensus. Effectively protect personal information during the data sharing process, and realize the full process of recording, verification, traceability, and auditing.

"Dandelion" artificial intelligence governance open platform

Monoun 2022 World Artificial Intelligence Conference

In order to solve the problems of difficult implementation of rules, separation of hard and hard research, and fragmentation of the governance, to ensure the healthy and sustainable development of artificial intelligence and health. "Open Platform of Dandelion Artificial Intelligence Governance -openeglab.

At present, there are still rare systems for governance and systematic governance systems in the world. The latest "Dandelion" OpenEglab, the goal is to create a system, practical artificial intelligence ethics and governance infrastructure, and explore the "rule-technology-scene-evaluation" integrated artificial intelligence governance system.

The platform currently supports interactive analysis of 36 common visual and language models, and is committed to the research of artificial intelligence governance that is closely integrated and promoted to promote the tight combination of exploration rules and technologies. The evaluation framework aims to conduct multi -dimensional evaluation of common artificial intelligence models.

Professor Xue Lan, director of the National Committee of the National New Generation of Artificial Intelligence Governance, said: "I hope to continue to accumulate innovative results and tools on the platform, promote the implementation of the concept of artificial intelligence governance, serve the efficiency of artificial intelligence governance, promote artificial intelligence, credible, credible, credible, credible, credible, credible, credible, credible, credible, and credible, credible, credible, credible, and to promote artificial intelligence and credibility. sustainable development."

- END -